Validate runtime split sizes are non-negative

When split sizes come from a runtime tensor input (not an attribute),
they were not validated for non-negative values. Crafted negative
split sizes that sum to the correct total could pass the sum check
and create tensors with negative dimensions. The attribute path
already had this validation.

Added non-negative check in both CPU PrepareForCompute and CUDA
PrepareForComputeLocal.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: bmehta001

onnxruntime/core/providers/cpu/tensor/split.h

@@ -65,6 +65,9 @@ class SplitBase {
       if (split_size_sum == -1) {
         split_size_sum = std::accumulate(split_sizes.cbegin(), split_sizes.cend(), 0LL);
       }
+      ORT_RETURN_IF_NOT(std::all_of(split_sizes.cbegin(), split_sizes.cend(),
+                                    [](int64_t value) { return value >= 0; }),
+                        "Invalid value in 'split' input. All values must be >= 0");
       if (split_sizes.size() != static_cast<size_t>(num_outputs) || split_size_sum != split_dim_size)
         return ORT_MAKE_STATUS(ONNXRUNTIME, FAIL,
                                "Cannot split using values in 'split' attribute. Axis=", axis_,

onnxruntime/core/providers/cuda/tensor/split.cc

@@ -89,6 +89,9 @@ Status SplitKernel::PrepareForComputeLocal(const TensorShape& input_shape,
     if (split_size_sum == -1) {
       split_size_sum = std::accumulate(split_sizes.cbegin(), split_sizes.cend(), 0LL);
     }
+    ORT_RETURN_IF_NOT(std::all_of(split_sizes.cbegin(), split_sizes.cend(),
+                                  [](int64_t value) { return value >= 0; }),
+                      "Invalid value in 'split' input. All values must be >= 0");
     if (split_sizes.size() != static_cast<size_t>(num_outputs) || split_size_sum != split_dim_size) {
       return ORT_MAKE_STATUS(ONNXRUNTIME, FAIL,
                              "Cannot split using values in 'split' attribute. Axis=", axis_,